in reply to Re^5: Why reftype and blessed are making my life harder than it needs to be (XS/_reftype)
in thread Why reftype and blessed are making my life harder than it needs to be

It seems sad, though, that reftype() won't be patched to XSRETURN_NO, especially as that makes more sense. Also, why can't the Perl version be patched to at least return !!0 instead of explicit undef?

XSRETURN_NO is how you do return !!0 in XS.

And the answer I think is here: 

D:\dev\>perl -e"print defined !!0"
1

[download]

But Im hopeful that the p5p powers that be will be receptive of the idea of a _reftype() sub that returns PL_no instead of Pl_undef. (Which is what XSRETURN_NO and XSRETURN_UNDEF do.)

---
$world=~s/war/peace/g

  • Comment on Re^6: Why reftype and blessed are making my life harder than it needs to be (XS/_reftype)
  • Download Code

Replies are listed 'Best First'.
Re^7: Why reftype and blessed are making my life harder than it needs to be (XS/_reftype)
by dragonchild (Archbishop) on Feb 10, 2006 at 14:27 UTC
    That's a good thing, at least for reftype(). I can see wanting to return undef from blessed() because you can bless into the empty string. But, the empty string is the representation for one of the 8 Perl scalar types. Or, am I just completely confused?
    My criteria for good software:
    1. Does it work?
    2. Can someone else come in, make a change, and be reasonably certain no bugs were introduced?
[reply]

      No, there is no perl intrinsic type that doesn't have a true value. And there are more than eight. We have: HASH, ARRAY, SCALAR, CODE, GLOB, IO, REF, and FORMAT, and then there is an intrinsic type that doesnt have its own designation, that being qr// objects, which report themselves as SCALAR but are actually not SCALAR refs. Oh, and i think in later perls you also have VERSION, (or something like that) but I could be mistaken.

      Regarding making reftype() return PL_no is that one must assume that people are doing 

      if (defined(reftype($x))) { ... }

      [download]

      or variants thereof. At least thats what I would expect the objection would be to changing the return of reftype. Hence my idea of providing a _reftype().

      ---
      $world=~s/war/peace/g

[reply]
[d/l]

        Don't make a new name, change the behavior at 'use' time. For example, if a module version number of at least X is demanded, then backward compatability behavior would not be required (so the function with saner behavior would be exported).

        - tye        

[reply]
        Gah. Because it was done wrong, all future versions have to do it wrong because people have a hack around it being done wrong. That's sad.

        Suggestion: Don't call it _reftype(). Call it what_reftype_shoulda_done()

        My criteria for good software:
        1. Does it work?
        2. Can someone else come in, make a change, and be reasonably certain no bugs were introduced?
[reply]

In Section Meditations